The goal is (1) to explore the possibility of deeper level of modeling, namely, incorporating mitochondria into our current nonlinear oxygen model; (2) to test the robustness of estimating regional MVO2 in the presence of noise; and (3) to analyze the PET data using this model. The first goal requires better understanding of mitochondria distribution and kinetics. The second goal is essential for developing a new noninvasive technique in measuring regional oxygen consumption using PET and O-15 oxygen. With the PET scan, only the residue function is available so we need to find out how big the error will be if using only the residue function instead of both residue and outflow dilution curves. In the presence of noise when small regions of interest (ROI) are used, we need to find out what is the smallest ROI size we can use and how that will affect the estimates of model parameters, especially MVO2. Monte Carlo simulation will be used for this task.Different levels of noise will be added to simulated data and the model will be fitted to many curves at the same noise level. The third goal can be achieved by analyzing more PET data from our dog experiments.